VTA resnet18

import vta
from vta.testing import simulator
from tvm import rpc, autotvm
import tvm
from tvm import relay
from dataclasses import dataclass
from pathlib import Path
from PIL import Image
from tqdm import tqdm
import numpy as np
from tvm_book.data.classification import ImageFolderDataset

def preprocess_image(
        image: np.ndarray,
        size: tuple[int] = (224, 224),
        mean: tuple[float] = (0.485, 0.456, 0.406),
        std: tuple[float] = (0.229, 0.224, 0.225)
    ):
    im = Image.fromarray(image)
    im = im.resize((256, 256), Image.Resampling.BILINEAR)
    ori_H, ori_W = im.size
    H, W = size
    space_W, space_H = (ori_W - W)//2, (ori_H - H)//2
    im = im.crop((space_W, space_H, ori_W-space_W, ori_H-space_H))
    image = np.array(im, dtype="float32")
    im.close()
    image = image/256
    image -= mean
    image /= std
    return image.astype(np.float32)

@dataclass
class ImageNet:
    root: str
    size: tuple[int] = (224, 224)
    mean: tuple[float] = (0.485, 0.456, 0.406)
    std: tuple[float] = (0.229, 0.224, 0.225)

    def __post_init__(self):
        self.root = Path(self.root) # 数据根目录
        self.valset = ImageFolderDataset(f"{self.root}/val")
        self.trainset = ImageFolderDataset(f"{self.root}/train")

    def calibrateset(self, calibrate_num: int = 200):
        """用于 TVM 量化的校准数据集
        """
        for k, (data, label) in tqdm(enumerate(self.trainset)):
            if k >= calibrate_num:
                break
            image = preprocess_image(data, self.size, self.mean, self.std)
            images = np.expand_dims(image, 0)
            images = images.transpose((0, 3, 1, 2))
            yield {"data": images}


def load_model(input_shape):
    from torchvision.models import resnet18, ResNet18_Weights
    import torch
    model = resnet18(weights=ResNet18_Weights.IMAGENET1K_V1)
    data = torch.randn(*input_shape)
    model = torch.jit.trace(model.eval(), data)
    return model.eval()

input_shape = 1, 3, 224, 224
input_name = "data"
traced_model = load_model(input_shape)
mod, params = relay.frontend.from_pytorch(
    traced_model, 
    [(input_name, input_shape)], 
    # use_parser_friendly_name=True
)
dataset = ImageNet("/media/pc/data/lxw/home/data/datasets/ILSVRC/")
with tvm.transform.PassContext(opt_level=3):
    with relay.quantize.qconfig(
        skip_conv_layers=[],
        calibrate_mode="kl_divergence", 
        weight_scale="max",
        round_for_shift=True,
        # rounding="TONEAREST", # "UPWARD" or "TONEAREST"
        calibrate_skip_layers=[],
        skip_dense_layer=False,
    ):
        qmod = relay.quantize.quantize(mod, params, dataset.calibrateset(calibrate_num=200))

200it [00:11, 17.31it/s]

remote = rpc.LocalSession()
env = vta.get_env()
ctx = remote.ext_dev(0)
assert env.target.device_name == "vta"

with autotvm.tophub.context(env.target):
    with tvm.transform.PassContext(opt_level=3):
        assert env.BLOCK_IN == env.BLOCK_OUT
        relay_prog = vta.top.graph_pack(
            qmod["main"],
            env.BATCH,
            env.BLOCK_OUT,
            env.WGT_WIDTH,
            start_name="nn.max_pool2d",
            stop_name="nn.adaptive_avg_pool2d",
            device_annot=(env.TARGET == "intelfocl"),
        )
    with vta.build_config(
            opt_level=3,
            disabled_pass={"AlterOpLayout",
                           "tir.CommonSubexprElimTIR"}
        ):
            lib = relay.build(relay_prog,
                              target=env.target,
                              params=params)
    # 将 inference library 发送到远程 RPC 服务器
    temp = tempdir()
    lib.export_library(temp.relpath("graphlib.tar"))
    remote.upload(temp.relpath("graphlib.tar"))
    lib = remote.load_module("graphlib.tar")

WARNING:autotvm:One or more operators have not been tuned. Please tune your model for better performance. Use DEBUG logging level to see more details.
[21:17:30] /media/pc/data/lxw/ai/tvm/src/tir/transforms/arg_binder.cc:95: Warning: Trying to bind buffer to another one with lower alignment requirement  required_alignment=256, provided_alignment=64

---------------------------------------------------------------------------
InternalError                             Traceback (most recent call last)
/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb 单元格 4 line 1
      <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=3'>4</a>     relay_prog = vta.top.graph_pack(
      <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=4'>5</a>         qmod["main"],
      <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=5'>6</a>         env.BATCH,
   (...)
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=10'>11</a>         device_annot=(env.TARGET == "intelfocl"),
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=11'>12</a>     )
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=12'>13</a> with vta.build_config(
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=13'>14</a>         opt_level=3,
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=14'>15</a>         disabled_pass={"AlterOpLayout",
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=15'>16</a>                        "tir.CommonSubexprElimTIR"}
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=16'>17</a>     ):
---> <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=17'>18</a>         lib = relay.build(relay_prog,
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=18'>19</a>                           target=env.target,
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=19'>20</a>                           params=params)
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=20'>21</a> # 将 inference library 发送到远程 RPC 服务器
     <a href='vscode-notebook-cell://ssh-remote%2B10.16.11.3/media/pc/data/lxw/ai/tvm/xinetzone/tvm-book/doc/vta/resnet18.ipynb#W3sdnNjb2RlLXJlbW90ZQ%3D%3D?line=21'>22</a> temp = tempdir()

File /media/pc/data/lxw/ai/tvm/xinetzone/__pypackages__/3.10/lib/tvm/relay/build_module.py:364, in build(ir_mod, target, target_host, executor, runtime, workspace_memory_pools, constant_memory_pools, params, mod_name)
    362 with tophub_context:
    363     bld_mod = BuildModule()
--> 364     graph_json, runtime_mod, params = bld_mod.build(
    365         mod=ir_mod,
    366         target=raw_targets,
    367         params=params,
    368         executor=executor,
    369         runtime=runtime,
    370         workspace_memory_pools=workspace_memory_pools,
    371         constant_memory_pools=constant_memory_pools,
    372         mod_name=mod_name,
    373     )
    374     func_metadata = bld_mod.get_function_metadata()
    375     devices = bld_mod.get_devices()

File /media/pc/data/lxw/ai/tvm/xinetzone/__pypackages__/3.10/lib/tvm/relay/build_module.py:161, in BuildModule.build(self, mod, target, target_host, executor, runtime, workspace_memory_pools, constant_memory_pools, params, mod_name)
    155 autotvm.GLOBAL_SCOPE.silent = (
    156     is_auto_scheduler_enabled() or is_meta_schedule_enabled() or old_autotvm_silent
    157 )
    159 mod_name = mangle_module_name(mod_name)
--> 161 self._build(
    162     mod,
    163     target,
    164     target_host,
    165     executor,
    166     runtime,
    167     workspace_memory_pools,
    168     constant_memory_pools,
    169     mod_name,
    170 )
    171 autotvm.GLOBAL_SCOPE.silent = old_autotvm_silent
    173 # Get artifacts

File /media/pc/data/lxw/ai/tvm/xinetzone/__pypackages__/3.10/lib/tvm/_ffi/_ctypes/packed_func.py:239, in PackedFuncBase.__call__(self, *args)
    227 ret_tcode = ctypes.c_int()
    228 if (
    229     _LIB.TVMFuncCall(
    230         self.handle,
   (...)
    237     != 0
    238 ):
--> 239     raise_last_ffi_error()
    240 _ = temp_args
    241 _ = args

File /media/pc/data/lxw/ai/tvm/xinetzone/__pypackages__/3.10/lib/tvm/_ffi/base.py:476, in raise_last_ffi_error()
    470 # The exception PyObject may contain a large amount of state,
    471 # including all stack frames that may be inspected in a later
    472 # PDB post-mortem.  Therefore, we must make sure to remove the
    473 # underlying PyObject* from the C++ side after we retrieve it.
    474 _LIB.TVMDropLastPythonError()
--> 476 raise py_err

InternalError: Traceback (most recent call last):
  146: tvm::relay::backend::RelayBuildModule::GetFunction(tvm::runtime::String const&, tvm::runtime::ObjectPtr<tvm::runtime::Object> const&)::{lambda(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)#3}::operator()(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
  145: tvm::relay::backend::RelayBuildModule::Build(tvm::IRModule, tvm::runtime::Array<tvm::Target, void> const&, tvm::Target const&, tvm::relay::Executor const&, tvm::relay::Runtime const&, tvm::WorkspaceMemoryPools const&, tvm::ConstantMemoryPools const&, tvm::runtime::String)
  144: tvm::relay::backend::RelayBuildModule::BuildRelay(tvm::IRModule, tvm::runtime::String const&)
  143: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::relay::backend::GraphExecutorCodegenModule::GetFunction(tvm::runtime::String const&, tvm::runtime::ObjectPtr<tvm::runtime::Object> const&)::{lambda(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)#2}> >::Call(tvm::runtime::PackedFuncObj const*, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)
  142: tvm::relay::backend::GraphExecutorCodegen::Codegen(tvm::IRModule, tvm::relay::Function, tvm::runtime::String)
  141: tvm::transform::Pass::operator()(tvm::IRModule) const
  140: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  139: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  138: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  137: tvm::transform::ModulePassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  136: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::runtime::TypedPackedFunc<tvm::IRModule (tvm::IRModule, tvm::transform::PassContext)>::AssignTypedLambda<tvm::relay::tec::LowerTE(tvm::runtime::String, tvm::CompilationConfig, std::function<void (tvm::BaseFunc)>)::{lambda(tvm::IRModule, tvm::transform::PassContext)#1}>(tvm::relay::tec::LowerTE(tvm::runtime::String, tvm::CompilationConfig, std::function<void (tvm::BaseFunc)>)::{lambda(tvm::IRModule, tvm::transform::PassContext)#1})::{lambda(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*)#1}> >::Call(tvm::runtime::PackedFuncObj const*, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)
  135: tvm::relay::tec::LowerTE(tvm::IRModule const&, tvm::runtime::String const&, std::function<void (tvm::BaseFunc)>, tvm::CompilationConfig)
  134: tvm::transform::Pass::operator()(tvm::IRModule) const
  133: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  132: tvm::relay::transform::FunctionPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  131: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::runtime::TypedPackedFunc<tvm::relay::Function (tvm::relay::Function, tvm::IRModule, tvm::transform::PassContext)>::AssignTypedLambda<tvm::relay::tec::LowerTensorExpr(tvm::relay::tec::TECompiler, std::function<void (tvm::BaseFunc)>, tvm::CompilationConfig)::{lambda(tvm::relay::Function, tvm::IRModule, tvm::transform::PassContext)#1}>(tvm::relay::tec::LowerTensorExpr(tvm::relay::tec::TECompiler, std::function<void (tvm::BaseFunc)>, tvm::CompilationConfig)::{lambda(tvm::relay::Function, tvm::IRModule, tvm::transform::PassContext)#1})::{lambda(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*)#1}> >::Call(tvm::runtime::PackedFuncObj const*, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)
  130: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  129: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  128: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::FunctionNode const*)
  127: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::FunctionNode const*)
  126: _ZN3tvm5relay9transform22DeviceAwareExprMutator21DeviceAwareVisit
  125: tvm::relay::ExprMutator::VisitExpr_(tvm::relay::FunctionNode const*)
  124: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  123: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  122: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  121: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  120: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  119: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  118: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  117: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  116: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  115: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  114: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  113: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  112: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  111: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  110: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  109: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  108: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  107: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  106: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  105: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  104: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  103: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  102: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  101: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  100: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  99: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  98: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  97: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  96: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  95: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  94: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  93: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  92: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  91: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  90: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  89: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  88: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  87: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  86: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  85: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  84: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  83: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  82: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  81: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  80: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  79: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  78: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  77: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  76: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  75: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  74: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  73: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  72: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  71: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  70: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  69: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  68: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  67: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  66: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  65: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  64: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  63: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  62: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  61: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  60: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  59: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  58: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  57: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  56: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  55: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  54: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  53: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  52: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  51: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  50: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  49: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  48: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  47: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  46: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  45: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  44: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  43: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  42: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  41: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  40: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  39: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  38: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  37: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  36: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  35: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  34: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  33: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  32: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  31: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  30: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  29: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  28: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  27: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  26: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  25: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  24: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  23: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  22: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  21: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  20: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  19: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  18: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  17: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  16: tvm::relay::ExprMutator::VisitExpr(tvm::RelayExpr const&)
  15: _ZZN3tvm5relay11ExprFunctorIFNS_9RelayExprERKS2_EE10InitVTableEvENUlRKNS_7r
  14: tvm::relay::transform::DeviceAwareExprMutator::VisitExpr_(tvm::relay::CallNode const*)
  13: tvm::relay::tec::LowerTensorExprMutator::DeviceAwareVisitExpr_(tvm::relay::CallNode const*)
  12: tvm::relay::tec::TECompilerImpl::Lower(tvm::relay::tec::CCacheKey const&)
  11: tvm::relay::tec::TECompilerImpl::LowerInternal(tvm::relay::tec::CCacheKey const&, tvm::GlobalVarSupply)
  10: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::te::Tensor, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
  9: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
  8: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
  7: tvm::transform::Pass::operator()(tvm::IRModule) const
  6: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  5: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  4: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  3: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
  2: tvm::runtime::PackedFuncObj::Extractor<tvm::runtime::PackedFuncSubObj<tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::AssignTypedLambda<tvm::tir::transform::StorageRewrite()::{lambda(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)#1}>(tvm::tir::transform::StorageRewrite()::{lambda(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)#1})::{lambda(tvm::runtime::TVMArgs const&, tvm::runtime::TVMRetValue*)#1}> >::Call(tvm::runtime::PackedFuncObj const*, tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*)
  1: tvm::tir::StoragePlanRewriter::PrepareNewAlloc()
  0: tvm::tir::StoragePlanRewriter::NewAllocTagMerged(tvm::tir::StoragePlanRewriter::StorageEntry*)
  File "/media/pc/data/lxw/ai/tvm/src/tir/transforms/storage_rewrite.cc", line 762
InternalError: Check failed: total_bits <= info->max_num_bits (7342080 vs. 1048576) : Allocation exceed bound of memory tag local.acc_buffer

print(relay_prog)

# 下载 ImageNet categories
categ_url = "https://github.com/uwsampl/web-data/raw/main/vta/models"
categ_fn = "synset.txt"
download.download(f"{categ_url}/{categ_fn}", categ_fn)
synset = eval(open(categ_fn).read())
# 下载测试图片
image_url = "https://homes.cs.washington.edu/~moreau/media/vta/cat.jpg"
image_fn = "cat.png"
download.download(image_url, image_fn)
# 准备用于推理的测试图像
image = Image.open(image_fn).resize((224, 224))
plt.imshow(image)
plt.show()
image = np.array(image) - np.array([123.0, 117.0, 104.0])
image /= np.array([58.395, 57.12, 57.375])
image = image.transpose((2, 0, 1))
image = image[np.newaxis, :]
image = np.repeat(image, env.BATCH, axis=0)

# 生成图执行器（graph executor） `m`。
m = graph_executor.GraphModule(lib["default"](ctxes))
# 设置网络参数和输入
m.set_input(**params)
m.set_input("data", image)